Soil moisture is predicted with spectroscopy based on the mechanism of moisture affecting characteristics of spectral reflectance of the soil, but most studies took reflectance as an independent variable in moisture predicting models, and paid only a little attention to absorption characteristics. In n this study 8 samples of black soil different in soil organic matter content were collected from its experiment field and prepared them into 102 soil samples using a new soil moisture adjusting method. The samples were then put individually into wide round glass disks. Spectral reflectances of the samples in the visible and near infrared region were measured with an ASD Field Spectroradiometer in the laboratory, yielding 10 spectral curves for each sample, of which a mean was worked out as the actual reflectance of the sample. Since the spectrometer responds unevenly to electromagnetic waves different in wavelength, spectral data need to be pre-processed for smoothing at a regular wavelength interval of 5 nm to diminish noise before data analysis. As the soil samples did not vary much in spectral characteristics, the continuum removal method was used to effectively make the characteristics of spectral adsorption and reflection prominent in the spectral curves. Soil spectral reflectance is comprehensive representation of soil physical and chemical parameters, and hence very sensitive to changes in soil organic matter (SOM) soil moisture, Fe, coarseness, mechanical composition and so on. However, the characteristic parameters of spectral adsorption valleys extracted with the continuum removal method reduced the sensitivities. The continuum removal method was applied with the aid of Software ENVI 4.6. The characteristic parameters of soil spectral adsorption that need to be extracted encompass area, depth and width of a spectral absorption valley. Correlation analysis was used to determine relationships of moisture content of the black soil with reflectance, spectral characteristic parameters and post-continuum-removal values. Based on the Simple Linear Regression, Stepwise Multiple Linear Regression(SMLR) and Partial Least Squares Regression(PLSR) method separately, high-spectrum models for prediction of black soil moisture content were built up using spectral reflectance, post-continuum-removal values and spectral adsorption characteristic parameters as independent variables, and moisture as dependent variable. Determination coefficient and RMSE were used to evaluate prediction accuracy of the models.. The higher the R² and the more stable and accurate the model and the lower the RMSE. Results show that (1) the soil spectral curve of Black soil has five spectral absorption valleys located at 510, 615, 1420, 1920 and 2210 nm, separately; and in predicting soil soil moisture content in black soil, spectral characteristic parameters are higher in correlativity than spectral reflectance, especially at 1420 and 1920 nm; (2) the models based on Simple Linear Regression, Stepwise Multiple Linear Regression (SMLR) and Partial Least Squares Regression (PLSR), separately, are all applicable to prediction of soil moisture content in black soil; and (3) the model based on simple linear regression using the characteristic parameters of the spectral absorption valley at 1 920 nm as independent value is high in prediction accuracy and low in input volume, and hence can be used as the theoretical basis for developing instant soil moisture measuring instruments. The models established in this study are high in stability and accuracy, which may be attributed to their use of just one type of soil and the new soil moisture adjusting method. Therefore, it can be concluded that the soil moisture high-spectrum prediction model based on spectral adsorption characteristic parameters is high in accuracy and stability and can be used for instant prediction of soil water contents.